Improved adaptive fractional order differential method for medical image enhancement

Abstract

In this work, we propose a construction of a new 5x5 fractional order differential mask that uses sixteen directions of gradient operator and weights each pixel in the mask by the Euclidean distance from the center of the mask. Then, we apply this new mask to the Adaptive Fractional Differential Algorithm (AFDA). The AFDA allows the optimal fractional order of each pixel to be obtained using an adaptive function constructed based on the area feature of image. Experimental results for medical images, show that the AFDA with the new mask gives better image enhancement than the original AFDA. It makes edges clearer, preserving texture details and improving the contrast of medical images. Moreover, we also use the proposed mask to restore the noisy images which are corrupted by the Gaussian noise. We use the peak signal to noise ratio (PSNR) and the structural similarity index measure (SSIM) to evaluate the quality of the denoised images. Changing the values of the fractional orders ν allows adjusting the mask coefficients for each image according to it characteristics. The experiments provide that the proposed mask has an influence on preserving more texture detail than the common used denoising filters. In addition, the output images have no significant blurring which can be indicated by higher SSIM. We conclude that the proposed mask can improve the result visually and in terms of PSNR and SSIM efficiently.